Nikhil Francis Giji

• Worked on a Human Activity Recognition project that uses sensor readings to classify human activities. • Conducted exploratory data analysis to find the underlying data distribution and implemented T-SNE. • Trained the model using support vector machine and produced an accuracy of 96 percent.

Worked on improving monocular depth estimation pipeline.

• Worked on building registration pipeline for 3D reconstruction of intra oral scans. • Implemented pipeline for real time monocular depth prediction from videos using Intel MiDaS Architecture. • Was able to reduce time for predicting depth maps using tensorflow lite model of the MiDaS Architecture.